Golden Sun (Sub 4 Hours????)

TheOthin

Edit history:

TheOthin: 2011-05-26 05:01:28 am

TheOthin: 2011-05-26 05:00:49 am

TheOthin: 2011-05-26 05:00:48 am

TheOthin: 2011-05-26 05:00:28 am

TheOthin: 2011-05-25 08:26:54 pm

TheOthin: 2011-05-25 08:25:59 pm

TheOthin: 2011-05-25 08:21:27 pm

TheOthin: 2011-05-25 08:20:35 pm

2011-05-25 08:15:13 pm

That's the thing - sometimes the divergence happens after a multi-target attack, sometimes it happens at a completely different time. I can try reproducing it to test other factors if necessary.

What do you know about the effect of the file screen on the RNG? It could explain the earlier divergences, and it doesn't seem to vary too much, so maybe it's simple to control. Certainly, the discussion I've heard about RNG abuse in TLA for rare weapons and the like doesn't seem to regard any such factors as a concern, and I don't think they're too far off from the truth. On the other hand, maybe any such variables only affect RNs farther down the line, past the point most RNG abuse guides are concerned with.

This wasn't what I wanted to find out with this test run. I'm not sure how to feel about that.

Edit: Well, now it diverged again. My attack pattern previously had been:

Turn 1:
Sheba attacks
King Scorpion defends
Jenna summons Mars
Felix summons Ramses

Turn 2:
Sheba summons Mercury
King Scorpion defends
Jenna casts Fume
Felix casts Earthquake

Turn 3:
King Scorpion defends
Jenna unleashes Cannon
Sheba unleashes Iron
Felix unleashes Echo

Turn 4:
Sheba unleashes Fog
Jenna summons Mars
King Scorpion unleashes Desert Gasp
Felix summons Ramses

The fourth time, I made all the same moves on those turns, but it used Twin Shear on the fourth turn - as it had sometimes done previously. I'm convinced there's some pattern to this that can be controlled, but I don't understand it. In any case, it's worth seeing if anyone else can replicate it.

I also wonder about how regular attacking affects the RNs; I can't seem to find it explained anywhere. It's not random, and generally a bad idea at high levels, but having a character use a regular attack was the only way I could find to make it defend for all of the first three turns, so it has to be different from Psynergy and summons. But at least that, unlike this other stuff, is actually consistent.

That said, I'm not going to be bothering with the King Scorpion anymore for now, since despite an unwanted divergence, the battle went smoothly enough that I won with no characters downed. Not impossible with this route and no RNG abuse, based on my experience, but rare enough that I'm taking a win where I can get it. I'm sure I'll be spending plenty of time on later bosses to figure them out, though.

That's the thing - sometimes the divergence happens after a multi-target attack, sometimes it happens at a completely different time. I can try reproducing it to test other factors if necessary.

What do you know about the effect of the file screen on the RNG? It could explain the earlier divergences, and it doesn't seem to vary too much, so maybe it's simple to control. Certainly, the discussion I've heard about RNG abuse in TLA for rare weapons and the like doesn't seem to regard any such factors as a concern, and I don't think they're too far off from the truth. On the other hand, maybe any such variables only affect RNs farther down the line, past the point most RNG abuse guides are concerned with.

This wasn't what I wanted to find out with this test run. I'm not sure how to feel about that.

Edit: Well, now it diverged again. My attack pattern previously had been:

Turn 1:
Sheba attacks
King Scorpion defends
Jenna summons Mars
Felix summons Ramses

Turn 2:
Sheba summons Mercury
King Scorpion defends
Jenna casts Fume
Felix casts Earthquake

Turn 3:
King Scorpion defends
Jenna unleashes Cannon
Sheba unleashes Iron
Felix unleashes Echo

Turn 4:
Sheba unleashes Fog
Jenna summons Mars
King Scorpion unleashes Desert Gasp
Felix summons Ramses

The fourth time, I made all the same moves on those turns, but it used Twin Shear on the fourth turn - as it had sometimes done previously. I'm convinced there's some pattern to this that can be controlled, but I don't understand it. In any case, it's worth seeing if anyone else can replicate it.

I also wonder about how regular attacking affects the RNs; I can't seem to find it explained anywhere. It's not random, and generally a bad idea at high levels, but having a character use a regular attack was the only way I could find to make it defend for all of the first three turns, so it has to be different from Psynergy and summons. But at least that, unlike this other stuff, is actually consistent.

That said, I'm not going to be bothering with the King Scorpion anymore for now, since despite an unwanted divergence, the battle went smoothly enough that I won with no characters downed. Not impossible with this route and no RNG abuse, based on my experience, but rare enough that I'm taking a win where I can get it. I'm sure I'll be spending plenty of time on later bosses to figure them out, though.

TheOthin

2011-05-26 03:18:59 pm

On an unrelated note, I found the battle with Briggs to be considerably faster and easier than other attempts with the use of one Oil Drop and one Weasel's Claw to clear the second set of Sea Fighters, both obtained from Yampi Desert, and I don't think obtaining them took much or any ingame time. I can also imagine later battles being assisted by those items for similar reasons. Seems like getting attacking items is worth it in the long run for more battles than just Saturos.

With that in mind, I have to wonder about other items being potentially good rare drops to seek - healing items such as Water of Life and Mist Potions, and certain equipment such as Feathered Robes with the capability to aid a summon-based offense. I don't think a total minimalist run is the way to go - just as some Djinn are well worth the time in the long run, the bit of time required to fight a battle to obtain these items throughout a run seems like it would be worth it, possibly in both games. Of course, there are limits on how many of these items are worth the time to get them, but it's worth consideration outside of the one battle where it seems to have been considered so far. Unless anyone has thoughts to the contrary?

Also, regarding my previous posts, the battle with Briggs seemed to support my theories. Over a number of attempts following hard resets, he seemed to display limited variation with his actions starting on the second turn, even as the events of the first turn remained constant. Nothing really new, but more evidence that the RNG of The Lost Age can be controlled perfectly for a short time, then controlled less perfectly past that time, but still doesn't immediately become impossible or unreasonable to control.

On an unrelated note, I found the battle with Briggs to be considerably faster and easier than other attempts with the use of one Oil Drop and one Weasel's Claw to clear the second set of Sea Fighters, both obtained from Yampi Desert, and I don't think obtaining them took much or any ingame time. I can also imagine later battles being assisted by those items for similar reasons. Seems like getting attacking items is worth it in the long run for more battles than just Saturos.

With that in mind, I have to wonder about other items being potentially good rare drops to seek - healing items such as Water of Life and Mist Potions, and certain equipment such as Feathered Robes with the capability to aid a summon-based offense. I don't think a total minimalist run is the way to go - just as some Djinn are well worth the time in the long run, the bit of time required to fight a battle to obtain these items throughout a run seems like it would be worth it, possibly in both games. Of course, there are limits on how many of these items are worth the time to get them, but it's worth consideration outside of the one battle where it seems to have been considered so far. Unless anyone has thoughts to the contrary?

Also, regarding my previous posts, the battle with Briggs seemed to support my theories. Over a number of attempts following hard resets, he seemed to display limited variation with his actions starting on the second turn, even as the events of the first turn remained constant. Nothing really new, but more evidence that the RNG of The Lost Age can be controlled perfectly for a short time, then controlled less perfectly past that time, but still doesn't immediately become impossible or unreasonable to control.

kirkq

Edit history:

kirkq: 2011-05-28 07:08:04 am

2011-05-28 07:06:48 am

Die Hard 2013.

Scripting:

Must be this version of this emulator: http://code.google.com/p/vba-rerecording/ (any fairly new revision should be fine)
Acquire your own rom, TASVideos does not want anything to do with rom downloads.

Copy and paste the code in the TASVideos thread into a text file and save it as GoldenSun.lua

Tools -> Lua Scripting -> New Lua Script Window -> Browse -> Run

It's just a memory display script. The script probably doesn't work with GS2, though there is a chance it could.

Finding Your Own Addresses: http://tasvideos.org/MemorySearch.html
Some Lua Info: http://tasvideos.org/LuaScripting.html

I intend to improve those site pages in the future as they are obviously incomplete and less than satisfactory at the moment.

I may look into some of the inconsistent RNG behavior in the near future, no promises.

TheOthin

2011-05-28 11:21:58 am

Much appreciated.

So I saved the script twice - once in its original form, and once with Isaac's name replaced in both instances with Felix. I tried loading the second with The Lost Age, and even when I reached Kandorean Temple, it didn't seem to do anything. Then I tried loading the first script with Golden Sun, and it also didn't seem to show my coordinates anywhere, even when I reached Sol Sanctum and went out-of-bounds. However, before I reached that point, I still had the script on during the battle with Saturos and Menardi, and for some reason, Isaac wouldn't die. Any damage he took, even being defeated, was instantly healed, until I turned off the script.

Any idea what's going on?

kirkq

2011-05-29 12:07:23 am

Die Hard 2013.

The words in the file have nothing to do with the game, they are just telling you the user what is going on. Don't try the script in Lost Age, it most likely doesn't work at all.

The lua you should be using is here: http://tasvideos.org/forum/viewtopic.php?t=3409&postdays=0&postorder=asc&start=88

If Isaac's HP won't decrease, then a revision may have been made with VBA that I need to look into. Either that or you did something silly and broke the script.

The script should display and/or hack X position, Y position, Isaac's PP, and a "Step Counter"

If these are not displaying on the screen at all something is wrong. If the correct values will not display on the screen, tell me what you are seeing. If Isaac's HP is on the screen I may know what happened, but tell me exactly what is happening before I look into that.

TheOthin

Edit history:

TheOthin: 2011-05-29 10:14:53 am

TheOthin: 2011-05-29 08:07:53 am

TheOthin: 2011-05-29 08:01:05 am

TheOthin: 2011-05-29 07:51:50 am

TheOthin: 2011-05-29 07:43:39 am

TheOthin: 2011-05-29 07:39:30 am

TheOthin: 2011-05-29 07:39:29 am

TheOthin: 2011-05-29 07:38:07 am

TheOthin: 2011-05-29 06:47:16 am

2011-05-29 06:46:27 am

I copied that script exactly into a Notepad document, started up the emulator and loaded a file on which I had reached Sol Sanctum, activated the script using the instructions you gave earlier, and nothing whatsoever displayed on the screen other than what normally would.

I had been originally using the script you had posted later; from your descriptions of the scripts, it sounds like that's the one I should be using. Even trying that one again, it also didn't work.

Are you certain that the stuff in the code boxes, when copied exactly, is the correct script? Could you please verify this, and then give a screenshot of what is supposed to happen?

Edit: Okay, scratch that. It seems to be working on the November 2009 version of the emulator. I'm not sure what went wrong with the newer versions, but it looks like you were right; something was changed with more recent ones, while it works with this.

Edit2: Well, it's tracking my coordinates, but I moved to some of the coordinates you described after getting out-of-bounds in Sol Sanctum (incredibly easy, shouldn't be the slightest bit of trouble to replicate on cartridge) and couldn't enter any doors. Even moving around a bit within those coordinates didn't seem to have any effect - or moving to the reverse of the coordinates in case you had the coordinates as (y, x) rather than (x, y). Do you know what's going on here? Are you able to reliably find doors at those coordinates when you perform the glitch?

Edit3: As far as I can tell, the script works in The Lost Age for keeping track of y coordinates, but it can't track x coordinates; they're always listed as either 0 or 33. And unsurprisingly, it couldn't keep track of or control Felix's PP or prevent random encounters, although I assume at least the former wouldn't be too difficult to change.

I copied that script exactly into a Notepad document, started up the emulator and loaded a file on which I had reached Sol Sanctum, activated the script using the instructions you gave earlier, and nothing whatsoever displayed on the screen other than what normally would.

I had been originally using the script you had posted later; from your descriptions of the scripts, it sounds like that's the one I should be using. Even trying that one again, it also didn't work.

Are you certain that the stuff in the code boxes, when copied exactly, is the correct script? Could you please verify this, and then give a screenshot of what is supposed to happen?

Edit: Okay, scratch that. It seems to be working on the November 2009 version of the emulator. I'm not sure what went wrong with the newer versions, but it looks like you were right; something was changed with more recent ones, while it works with this.

Edit2: Well, it's tracking my coordinates, but I moved to some of the coordinates you described after getting out-of-bounds in Sol Sanctum (incredibly easy, shouldn't be the slightest bit of trouble to replicate on cartridge) and couldn't enter any doors. Even moving around a bit within those coordinates didn't seem to have any effect - or moving to the reverse of the coordinates in case you had the coordinates as (y, x) rather than (x, y). Do you know what's going on here? Are you able to reliably find doors at those coordinates when you perform the glitch?

Edit3: As far as I can tell, the script works in The Lost Age for keeping track of y coordinates, but it can't track x coordinates; they're always listed as either 0 or 33. And unsurprisingly, it couldn't keep track of or control Felix's PP or prevent random encounters, although I assume at least the former wouldn't be too difficult to change.

LinkTheValiant

Edit history:

LinkTheValiant: 2011-05-29 01:11:16 pm

2011-05-29 12:56:33 pm

May the way of the Hero lead to the Triforce

Alright, regarding RNG abuse, the guide I've been using is Link Kirby's guide found here. Unfortunately it's specifically devoted to abuse of the RNG for item drops, and as such involves no discussion of enemy attacks.

According to the FAQ, there are actually two different RNGs. One is the "general" RNG, which determines encounter composition, steps to next battle, NPC movement, and so on. This generator uses the button presses at the start screen as seeds. (At least partly, probably entirely.)

The other RNG, completely unaffected by the first, is the battle RNG, which so far as I know is used only in battle. This, as I understand it, determines enemy attack patterns and item drops. I'm fairly sure that it's this second one that's either glitched or not entirely working as LK thinks it does.

There's one other possibility that just occurred to me. Bosses have a different move pattern from normal encounters, as evidenced by the fact that they get actions in the first round after a hard reset. (Normal encounters do not, a fact which is exploited in item drop abuse.) This "enhanced moveset" may call the General RNG at specific points during a battle to change things up. Again, just a guess out of thin air, but something to think about if the divergence is consistently on the third or fourth round.

The problem is, I've never really poked around the guts of the RNG myself. I've simply used what LK already figured out and left it at that. I'm not particularly proficient with reading memory, so I'm not the one to figure out precisely what's going on with the generators. I'll keep poking around the Internet to see whether anyone knows, but for now that's all the information I really have.

EDIT: After some poking around, (okay, actually just a quick look at teh Google I should have taken before posting,) I found this. It's mostly more of the same from LK's guide, but a section stood out:

Quote:

Golden Sun uses the ANSI-C standard rand function as its Pseudo-Random Number
Generator. A variant of the source code can be seen here:

http://hpcf.nersc.gov/vendor_docs/ibm/libs/basetrf2/rand.htm

There is one difference: Golden Sun's RNG uses an unsigned 4-byte number and
returns it with no shifting, so you'd replace "return ((next >>16) & 32757);"
with "return (next & 4294967295);"

This, ordinarily, is a perfectly capable RNG, but there is one flaw with what
Golden Sun does with it.

It only makes a new random number when something random happens. What's more,
it uses *two* copies of the RNG to create two random numbers, and the
first RNG only handles battle situations.

As such, it is easy to control when a new random number is generated.

On top of this, the RNGs are seeded with 0 when the GBA is reset, and because
of the way the game has been programmed, the RNG controlling battle events will
*continue* to be 0 even when you load the game...

Something more to think about, if nothing else. I still can't figure out how the General RNG figures in though.

Alright, regarding RNG abuse, the guide I've been using is Link Kirby's guide found [url=http://dw.com.com/redir?asid=0&astid=8&siteid=19&edid=107&destCat=&destURL=http%3A%2F%2Fdb.gamefaqs.com%2Fportable%2Fgbadvance%2Ffile%2Fgolden_sun_2_random.txt]here[/url]. Unfortunately it's specifically devoted to abuse of the RNG for item drops, and as such involves no discussion of enemy attacks.

According to the FAQ, there are actually two different RNGs. One is the "general" RNG, which determines encounter composition, steps to next battle, NPC movement, and so on. This generator uses the button presses at the start screen as seeds. (At least partly, probably entirely.)

The other RNG, completely unaffected by the first, is the battle RNG, which so far as I know is used only in battle. This, as I understand it, determines enemy attack patterns and item drops. I'm fairly sure that it's this second one that's either glitched or not entirely working as LK thinks it does.

There's one other possibility that just occurred to me. Bosses have a different move pattern from normal encounters, as evidenced by the fact that they get actions in the first round after a hard reset. (Normal encounters do not, a fact which is exploited in item drop abuse.) This "enhanced moveset" may call the General RNG at specific points during a battle to change things up. Again, just a guess out of thin air, but something to think about if the divergence is consistently on the third or fourth round.

The problem is, I've never really poked around the guts of the RNG myself. I've simply used what LK already figured out and left it at that. I'm not particularly proficient with reading memory, so I'm not the one to figure out precisely what's going on with the generators. I'll keep poking around the Internet to see whether anyone knows, but for now that's all the information I really have.

EDIT: After some poking around, (okay, actually just a quick look at teh Google I should have taken before posting,) I found [url=http://www.goldensunrealm.com/gs/nickpresta_rng-guide.txt]this[/url]. It's mostly more of the same from LK's guide, but a section stood out:

[quote]Golden Sun uses the ANSI-C standard rand function as its Pseudo-Random Number
Generator. A variant of the source code can be seen here:

http://hpcf.nersc.gov/vendor_docs/ibm/libs/basetrf2/rand.htm

There is one difference: Golden Sun's RNG uses an unsigned 4-byte number and
returns it with no shifting, so you'd replace "return ((next >>16) & 32757);"
with "return (next & 4294967295);"

This, ordinarily, is a perfectly capable RNG, but there is one flaw with what
Golden Sun does with it.

It only makes a new random number when something random happens. What's more, 
it uses *two* copies of the RNG to create two random numbers, and the 
first RNG only handles battle situations.

As such, it is easy to control when a new random number is generated.

On top of this, the RNGs are seeded with 0 when the GBA is reset, and because 
of the way the game has been programmed, the RNG controlling battle events will
*continue* to be 0 even when you load the game...[/quote]

Something more to think about, if nothing else. I still can't figure out how the General RNG figures in though.

TheOthin

Edit history:

TheOthin: 2011-05-29 06:23:40 pm

TheOthin: 2011-05-29 05:46:06 pm

TheOthin: 2011-05-29 05:45:18 pm

TheOthin: 2011-05-29 02:42:27 pm

TheOthin: 2011-05-29 02:41:56 pm

TheOthin: 2011-05-29 02:15:27 pm

2011-05-29 02:13:47 pm

If the General RNG influences targets, it could explain some things... I still don't know what to make of the divergences in certain actions even before using any multi-target moves, though.

Has anyone else been able to replicate that and notice any possible explanation in doing so?

Edit: So I went back to mess around with Sol Sanctum again. I decided to try for the door that leads just halfway through Sol Sanctum, where you start a Kraden cutscene from inside a pillar, which I had been able to reach on the cartridge. This one was right where I expected it to be. Not sure what the implications of this are.

Edit2: I decided to see if the code could easily be adjusted to work for The Lost Age. I couldn't fix Felix's PP or the random encounters, but I was able to make it keep track of the x coordinate. The original script seems to look at the variable 0x02030ec4 or something to keep track of the x coordinate in Golden Sun; changing it to 0x02030ec8 made it work for The Lost Age. Or whatever all the code and stuff means. I don't understand the mechanics behind any of this, but just by making that change, the code can be used to track x and y coordinates in the second game instead. I am beginning a search for important doors in The Lost Age using this immediately. Hopefully they'll be more reliable in being found.

Edit3: Trying the code on Gondowan Cliffs, the script would sometimes lose track of both coordinates, then compensate later. I'm not sure whether or not this is proof of a flaw in the script's compatibility with The Lost Age, since based on the camera movements I observed both on the emulator and trying the glitch in the same area ingame, the same thing is going on in the game itself.

If the General RNG influences targets, it could explain some things... I still don't know what to make of the divergences in certain actions even before using any multi-target moves, though.

Has anyone else been able to replicate that and notice any possible explanation in doing so?

Edit: So I went back to mess around with Sol Sanctum again. I decided to try for the door that leads just halfway through Sol Sanctum, where you start a Kraden cutscene from inside a pillar, which I [i]had[/i] been able to reach on the cartridge. This one was right where I expected it to be. Not sure what the implications of this are.

Edit2: I decided to see if the code could easily be adjusted to work for The Lost Age. I couldn't fix Felix's PP or the random encounters, but I was able to make it keep track of the x coordinate. The original script seems to look at the variable 0x02030ec4 or something to keep track of the x coordinate in Golden Sun; changing it to 0x02030ec8 made it work for The Lost Age. Or whatever all the code and stuff means. I don't understand the mechanics behind any of this, but just by making that change, the code can be used to track x and y coordinates in the second game instead. I am beginning a search for important doors in The Lost Age using this immediately. Hopefully they'll be more reliable in being found.

Edit3: Trying the code on Gondowan Cliffs, the script would sometimes lose track of both coordinates, then compensate later. I'm not sure whether or not this is proof of a flaw in the script's compatibility with The Lost Age, since based on the camera movements I observed both on the emulator and trying the glitch in the same area ingame, the same thing is going on in the game itself.

Manocheese

2011-05-29 07:27:38 pm

Yes, a cucco riding the ground.

Gondowan Cliffs might have an issue there because of the way the camera suddenly shifts. I've played around out of bounds there and I've gotten some weird stuff to happen. I don't think any other area has something like that.

TheOthin

2011-05-29 07:44:41 pm

Quote from Manocheese:

Gondowan Cliffs might have an issue there because of the way the camera suddenly shifts. I've played around out of bounds there and I've gotten some weird stuff to happen. I don't think any other area has something like that.

That's what I mean. The displayed coordinates change the same way the camera does, and there are a bunch of weird walls I don't know how to explain.

On the note of weird stuff, I once managed to escape all those walls and get really far away from the main area on cartridge, only to somehow find myself continuously bouncing straight down the screen and getting stuck in an infinite loop, returning to the minimum coordinates as soon as the maximum coordinates were reached and just continuing to bounce. Oddly enough, I could still access menus and use Psynergy, although the bouncing never stopped... I eventually used Retreat to escape. I've been trying to replicate that on the emulator so I can find coordinates that trigger it, but no luck so far. Was that one of the things you've seen?

It's as if the area is built to guard against what may be the biggest sequence break possible in the series.

[quote="Manocheese"]Gondowan Cliffs might have an issue there because of the way the camera suddenly shifts. I've played around out of bounds there and I've gotten some weird stuff to happen. I don't think any other area has something like that.[/quote]
That's what I mean. The displayed coordinates change the same way the camera does, and there are a bunch of weird walls I don't know how to explain.

On the note of weird stuff, I once managed to escape all those walls and get really far away from the main area on cartridge, only to somehow find myself continuously bouncing straight down the screen and getting stuck in an infinite loop, returning to the minimum coordinates as soon as the maximum coordinates were reached and just continuing to bounce. Oddly enough, I could still access menus and use Psynergy, although the bouncing never stopped... I eventually used Retreat to escape. I've been trying to replicate that on the emulator so I can find coordinates that trigger it, but no luck so far. Was that one of the things you've seen?

It's as if the area is built to guard against what may be the biggest sequence break possible in the series.

kirkq

2011-05-30 12:00:54 am

Die Hard 2013.

Quote from TheOthin:

Okay, scratch that. It seems to be working on the November 2009 version of the emulator. I'm not sure what went wrong with the newer versions, but it looks like you were right; something was changed with more recent ones, while it works with this.

Thanks for noting that, I'll look into it later.

Quote from TheOthin:

Edit2: Well, it's tracking my coordinates, but I moved to some of the coordinates you described after getting out-of-bounds in Sol Sanctum (incredibly easy, shouldn't be the slightest bit of trouble to replicate on cartridge) and couldn't enter any doors. Even moving around a bit within those coordinates didn't seem to have any effect - or moving to the reverse of the coordinates in case you had the coordinates as (y, x) rather than (x, y). Do you know what's going on here? Are you able to reliably find doors at those coordinates when you perform the glitch?

You should be able to get in them fairly easily if you move around a bit. If I recall they are (X,Y), but I haven't used it in a while. It's also mildly possible that an emulation error in older versions was producing the doors. Have you found any of the doors at any coordinates?

If this doesn't get sorted out I'll look at it in a day or two. I didn't expect any of these problems to occur.

[quote="TheOthin"]Okay, scratch that. It seems to be working on the November 2009 version of the emulator. I'm not sure what went wrong with the newer versions, but it looks like you were right; something was changed with more recent ones, while it works with this.[/quote]

Thanks for noting that, I'll look into it later.

[quote="TheOthin"]Edit2: Well, it's tracking my coordinates, but I moved to some of the coordinates you described after getting out-of-bounds in Sol Sanctum (incredibly easy, shouldn't be the slightest bit of trouble to replicate on cartridge) and couldn't enter any doors. Even moving around a bit within those coordinates didn't seem to have any effect - or moving to the reverse of the coordinates in case you had the coordinates as (y, x) rather than (x, y). Do you know what's going on here? Are you able to reliably find doors at those coordinates when you perform the glitch?[/quote]

You should be able to get in them fairly easily if you move around a bit.  If I recall they are (X,Y), but I haven't used it in a while.  It's also mildly possible that an emulation error in older versions was producing the doors.  Have you found any of the doors at any coordinates?

If this doesn't get sorted out I'll look at it in a day or two.  I didn't expect any of these problems to occur.

TheOthin

Edit history:

TheOthin: 2011-05-30 07:28:12 am

TheOthin: 2011-05-30 07:22:59 am

TheOthin: 2011-05-30 07:19:50 am

TheOthin: 2011-05-30 06:33:09 am

TheOthin: 2011-05-30 06:16:33 am

TheOthin: 2011-05-30 06:16:21 am

TheOthin: 2011-05-30 06:07:59 am

TheOthin: 2011-05-30 06:07:36 am

TheOthin: 2011-05-30 04:55:15 am

TheOthin: 2011-05-30 04:54:50 am

2011-05-30 04:54:11 am

On the emulator, I've tried to reach most of the five doors you listed the coordinates to in the TASVideos thread, and couldn't get into any of them. I've tried to reach the same doors on the cartridge, but also couldn't; I'm not sure whether that was because of the doors not being there or simply failing to reach them. On both emulator and cartridge, I've been able to get through the other door in the room you fall into to get halfway through Sol Sanctum, but never any of the glitched doors requiring underflowing the coordinates.

The next thing I'm going to try is using the emulator and its ability to track coordinates as a guide for the cartridge, played simultaneously, to reach those coordinates to verify whether or not there really is anything there. I also might try looking into Goma Cave; do you have the coordinates you used there?

Edit: Using this method, I was able to synchronize movements on cartridge with those on emulator, which I verified by having both enter the visible area in the same place at the same time, starting from about the same place out-of-bounds from the third room. From there, I moved to each of the five sets of coordinates where you described finding doors, and found nothing at any of them either on cartridge or on emulator. I also checked all eight adjacent coordinates for all of those doors on both cartridge and on emulator, and there was nothing there, either.

There are definitely ways to use this glitch to fall into other rooms and create shortcuts. And there are definitely weird things that can happen with this glitch if you go far out of bounds and underflow the game's coordinates, including glitched doors. Having verified both on the cartridge, I have no doubt that save retreat glitching has effective uses on the cartridge, both for speedruns and for non-speed runs. That said, I don't know the coordinates of the things I was able to find out of bounds, and none of them seem useful - taking damage for no reason, being forced out of the area, and that one weird infinite loop thing are the only things I remember. I don't know what to make of the doors you found, but to be on the safe side, I recommend assuming for now that any doors found on an emulator are errors until replicated on the cartridge - preferably multiple times, by multiple people, to make sure the doors actually have fixed coordinates in the first place. It may also be wise for us to test multiple versions of the emulator to see if we can find any differences attributable to emulation error, so we know how to regard those differences and which versions seem more reliable.

Edit2: Forgot to mention this. After I couldn't find the doors, I started walking around in random directions on the cartridge to see if I could find them somewhere else near where I was. I made the mistake of not tracking my movements with the emulator anymore. Eventually, after walking quite a ways up, I hit a door that took me to the second room in Sol Sanctum. Not at all useful, but I don't know what the implications are. I'm going to try to find that door again and record its coordinates, maybe we can figure something out based on them.

Edit3: Decided to first scour the area with the emulator and its turbo function, so that I could go back with the cartridge if I found anything. Running as far as I could left or right for each y value from 4275 to 4294, I found some weird walls, but no doors anywhere in the area. Unless they're between coordinates or something, I don't think there are any doors to be found there - not consistently, anyway. I might have missed a few potential areas to look at since I usually just ran in one direction, moved up one coordinate, and ran in the other direction, since the space between the walls seemed to get wider each time (eventually disappearing entirely), but the farthest left I could run was generally around 2000 and the farthest right I could run was generally around 1000, so anything I might have missed wasn't anywhere close to the area where we were originally looking, unless I did something wrong I wasn't thinking of. I'd say that when examining this glitch, the vagueness of it means that any tests should be repeated by someone else to make sure, but it's probably worth mentioning that even with the turbo function, this took a while. Probably better to look at a more limited area if anyone's actually interested.

On the emulator, I've tried to reach most of the five doors you listed the coordinates to in the TASVideos thread, and couldn't get into any of them. I've tried to reach the same doors on the cartridge, but also couldn't; I'm not sure whether that was because of the doors not being there or simply failing to reach them. On both emulator and cartridge, I've been able to get through the other door in the room you fall into to get halfway through Sol Sanctum, but never any of the glitched doors requiring underflowing the coordinates.

The next thing I'm going to try is using the emulator and its ability to track coordinates as a guide for the cartridge, played simultaneously, to reach those coordinates to verify whether or not there really is anything there. I also might try looking into Goma Cave; do you have the coordinates you used there?

Edit: Using this method, I was able to synchronize movements on cartridge with those on emulator, which I verified by having both enter the visible area in the same place at the same time, starting from about the same place out-of-bounds from the third room. From there, I moved to each of the five sets of coordinates where you described finding doors, and found nothing at any of them either on cartridge or on emulator. I also checked all eight adjacent coordinates for all of those doors on both cartridge and on emulator, and there was nothing there, either.

There are definitely ways to use this glitch to fall into other rooms and create shortcuts. And there are definitely weird things that can happen with this glitch if you go far out of bounds and underflow the game's coordinates, including glitched doors. Having verified both on the cartridge, I have no doubt that save retreat glitching has effective uses on the cartridge, both for speedruns and for non-speed runs. That said, I don't know the coordinates of the things I was able to find out of bounds, and none of them seem useful - taking damage for no reason, being forced out of the area, and that one weird infinite loop thing are the only things I remember. I don't know what to make of the doors you found, but to be on the safe side, I recommend assuming for now that any doors found on an emulator are errors until replicated on the cartridge - preferably multiple times, by multiple people, to make sure the doors actually have fixed coordinates in the first place. It may also be wise for us to test multiple versions of the emulator to see if we can find any differences attributable to emulation error, so we know how to regard those differences and which versions seem more reliable.

Edit2: Forgot to mention this. After I couldn't find the doors, I started walking around in random directions on the cartridge to see if I could find them somewhere else near where I was. I made the mistake of not tracking my movements with the emulator anymore. Eventually, after walking quite a ways up, I hit a door that took me to the second room in Sol Sanctum. Not at all useful, but I don't know what the implications are. I'm going to try to find that door again and record its coordinates, maybe we can figure something out based on them.

Edit3: Decided to first scour the area with the emulator and its turbo function, so that I could go back with the cartridge if I found anything. Running as far as I could left or right for each y value from 4275 to 4294, I found some weird walls, but no doors anywhere in the area. Unless they're between coordinates or something, I don't think there are any doors to be found there - not consistently, anyway. I might have missed a few potential areas to look at since I usually just ran in one direction, moved up one coordinate, and ran in the other direction, since the space between the walls seemed to get wider each time (eventually disappearing entirely), but the farthest left I could run was generally around 2000 and the farthest right I could run was generally around 1000, so anything I might have missed wasn't anywhere close to the area where we were originally looking, unless I did something wrong I wasn't thinking of. I'd say that when examining this glitch, the vagueness of it means that any tests should be repeated by someone else to make sure, but it's probably worth mentioning that even with the turbo function, this took a while. Probably better to look at a more limited area if anyone's actually interested.

kirkq

2011-05-30 09:41:19 am

Die Hard 2013.

Hitting _ANY_ door on emulator confirms that these shortcuts still exist somewhere. Now the question is if VBA is emulating it correctly now even. (There have been a lot of updates to VBA in the last year or so, so I am guessing something highly relevant was changed or corrected.) It is possible my original glitched doors were an emulation error. We need to find which version of the emulator, if any, are reliable. (My guess is that there was only one or two fundamental, relevant changes in all of the emulator revisions.) Thanks for clarifying all of this.

TheOthin

Edit history:

TheOthin: 2011-05-30 10:40:14 am

2011-05-30 10:38:31 am

Do you know which version of the emulator you were using when you found those doors? I went with the November 2009 one because it was the most recent one as of the post where you gave the script and coordinates, and therefore seemed like the one you would have been using, but if emulator version differences are a problem, it sounds like that isn't the case.

Manocheese

2011-05-30 12:07:51 pm

Yes, a cucco riding the ground.

Quote from TheOthin:

On the note of weird stuff, I once managed to escape all those walls and get really far away from the main area on cartridge, only to somehow find myself continuously bouncing straight down the screen and getting stuck in an infinite loop, returning to the minimum coordinates as soon as the maximum coordinates were reached and just continuing to bounce. Oddly enough, I could still access menus and use Psynergy, although the bouncing never stopped... I eventually used Retreat to escape. I've been trying to replicate that on the emulator so I can find coordinates that trigger it, but no luck so far. Was that one of the things you've seen?

I've mainly seen graphical glitches. If you enter from the Madra side, go to the Kibombo warrior room, and do the save Retreat glitch, you'll end up somewhere left of the area. Walk down and right and you'll see some weird crap that doesn't belong there. There's an icy cliff that looks like it's from Mars Lighthouse and some roots that look like something from Kolima Forest. I think that these are caused by the camera shift, although even weirder graphical things can happen--play around in Shaman Village for a while and you'll find numerous entrances to what I call "Glitchy Road", a screwed up version of Trial Road.

Can this script you guys are using be made to work on the world map in TLA? You may have seen some videos I made of a glitch that allows you to get OoB on the world map. There are lots of loading zones for various places out there, but I haven't found any useful ones. Perhaps the script could help. Here are the videos:

EDIT (not really): I was just playing around OoB in Gondowan Cliffs and I found a loading zone to the outside (Madra side). It was somewhere north of the visible portion of the area. Perhaps there's a loading zone to the Kibombo side; that may allow us to skip the first trip to Alhafra. I'll keep testing on console. You may have more luck with the script, though.

[quote="TheOthin"]On the note of weird stuff, I once managed to escape all those walls and get really far away from the main area on cartridge, only to somehow find myself continuously bouncing straight down the screen and getting stuck in an infinite loop, returning to the minimum coordinates as soon as the maximum coordinates were reached and just continuing to bounce. Oddly enough, I could still access menus and use Psynergy, although the bouncing never stopped... I eventually used Retreat to escape. I've been trying to replicate that on the emulator so I can find coordinates that trigger it, but no luck so far. Was that one of the things you've seen?[/quote]

I've mainly seen graphical glitches. If you enter from the Madra side, go to the Kibombo warrior room, and do the save Retreat glitch, you'll end up somewhere left of the area. Walk down and right and you'll see some weird crap that doesn't belong there. There's an icy cliff that looks like it's from Mars Lighthouse and some roots that look like something from Kolima Forest. I think that these are caused by the camera shift, although even weirder graphical things can happen--play around in Shaman Village for a while and you'll find numerous entrances to what I call "Glitchy Road", a screwed up version of Trial Road.

Can this script you guys are using be made to work on the world map in TLA? You may have seen some videos I made of a glitch that allows you to get OoB on the world map. There are lots of loading zones for various places out there, but I haven't found any useful ones. Perhaps the script could help. Here are the videos:

http://www.youtube.com/watch?v=Dy_Xq75tsaM
http://www.youtube.com/watch?v=xqOIrwoK5r8
http://www.youtube.com/watch?v=yJczBBACm-s
http://www.youtube.com/watch?v=xEngS-g-Vb4
http://www.youtube.com/watch?v=piHX27zfSlY

EDIT (not really): I was just playing around OoB in Gondowan Cliffs and I found a loading zone to the outside (Madra side). It was somewhere north of the visible portion of the area. Perhaps there's a loading zone to the Kibombo side; that may allow us to skip the first trip to Alhafra. I'll keep testing on console. You may have more luck with the script, though.

TheOthin

2011-05-30 12:18:17 pm

The current version of the script only works for ingame maps; on the overworld, both values seem to be fixed at 4024. I tried changing the last letter... digit... character... thing of the coding part, since it had worked before, but I couldn't get anything that changed with overworld movement.

Manocheese

Edit history:

Manocheese: 2011-05-30 03:32:49 pm

Manocheese: 2011-05-30 03:01:23 pm

2011-05-30 02:38:09 pm

Yes, a cucco riding the ground.

I was able to use the save Retreat glitch to get to the Kibombo side of the world map from Gondowan Cliffs using the method described in my last post. That means we can get to Gondowan early. I've gotten to the loading zone twice, but I haven't found a consistent way to do so. When I do, I'll post a video.

EDIT: It seems that the area outside of the visible map is somewhat random. If you keep soft resetting, the contents change. Turning the power off and on seems to reset them. I'll keep looking for a consistent way to find the loading zone. I've gotten to it three times now.

EDIT2: I found a somewhat consistent method:

FlamingMage

2011-06-10 01:18:23 pm

Lol guys, get a thread Tongue

So, is the general consensus that these doors are too difficult to find without an emulator for a console speedrun to use them? Or should I be learning to find them?

Manocheese

2011-06-10 01:52:28 pm

Yes, a cucco riding the ground.

I found the Gondowan early door on console. It's probably more efficient to use the script on emulator, though.

FlamingMage

2011-06-10 10:27:09 pm

Thanks manocheese Smiley

I guess I'm primarily talking about the Sol Sanctum door that Othin was having trouble reaching. If these doors are within the range of a console player, I guess I have to rework my whole route Sad

kirkq

Edit history:

kirkq: 2011-06-11 08:46:45 pm

kirkq: 2011-06-11 08:40:56 pm

kirkq: 2011-06-11 08:39:02 pm

kirkq: 2011-06-11 08:00:07 pm

2011-06-11 07:55:58 pm

Die Hard 2013.

I tested four emulator versions from VBA 22 to VBA 24. They all acted the same. I then tested some of my doors on the handheld and they acted the same. There should not be problems in the scripts, doors, or methods I've presented.

Sol Sanctum Door 1: 30, 4282

1: Go to room 3. Stand exactly even with the third pillar from the right, pushed up against it.
2: Do save retreat glitch.
3: Walk left for 3+ seconds (you will run into a wall and stop)
4: Walk up for 3+ seconds (you will run into a wall and stop)
5: Walk up+left for 4+ seconds (you will run into a wall and stop)
6: Walk right for 2+ seconds (you will run into a wall and stop)
7: Walk up-left for 2+ seconds

Sol Sanctum Door 6: 4289, 4284

1: Go to room 3. Stand exactly even with the third pillar from the right, pushed up against it.
2: Do save retreat glitch.
3: left 3+s, up 3+s, left 12+s, up 2+s, left 2+s, up 2+s

TLA Memory addresses to put in the script
X Position: 0x020322F4
Y Position: 0x020322FC
Felix PP: 0x02000A8A
Step Counter 0x0200049A
(Other Step Counter that isn't of any concern to the script 0x02000496)

All you need to change in the script is the addresses (things that start with 0x).

I tested four emulator versions from VBA 22 to VBA 24.  They all acted the same.  I then tested some of my doors on the handheld and they acted the same.  There should not be problems in the scripts, doors, or methods I've presented.

Sol Sanctum Door 1: 30, 4282

1: Go to room 3.  Stand exactly even with the third pillar from the right, pushed up against it.
2: Do save retreat glitch.
3: Walk left for 3+ seconds (you will run into a wall and stop)
4: Walk up for 3+ seconds (you will run into a wall and stop)
5: Walk up+left for 4+ seconds (you will run into a wall and stop)
6: Walk right for 2+ seconds (you will run into a wall and stop)
7: Walk up-left for 2+ seconds

Sol Sanctum Door 6: 4289, 4284

1: Go to room 3.  Stand exactly even with the third pillar from the right, pushed up against it.
2: Do save retreat glitch.
3: left 3+s, up 3+s, left 12+s, up 2+s, left 2+s, up 2+s

TLA Memory addresses to put in the script
X Position: 0x020322F4
Y Position: 0x020322FC
Felix PP: 0x02000A8A
Step Counter 0x0200049A
(Other Step Counter that isn't of any concern to the script 0x02000496)

All you need to change in the script is the addresses (things that start with 0x).

FlamingMage

2011-06-11 07:57:47 pm

Thanks kirkq, those descriptions look awesome, I'll try it out as soon as I'm back with my GBA Cheesy

FionordeQuester

2011-06-11 08:22:20 pm

FlamingMage, if you end up having this be a run with glitches, could you please post the route you would've used had you not discovered these glitches?

FlamingMage

2011-06-11 09:28:03 pm

I'll at least discuss the route in my commentaries, and probably can post it for you, sure.

kirkq

Edit history:

kirkq: 2011-06-15 01:15:43 pm

kirkq: 2011-06-15 01:15:24 pm

kirkq: 2011-06-15 05:03:18 am

kirkq: 2011-06-15 01:58:55 am

kirkq: 2011-06-14 04:40:04 pm

kirkq: 2011-06-14 11:14:14 am

kirkq: 2011-06-14 10:56:40 am

2011-06-14 10:56:09 am

Die Hard 2013.

I wrote a bot to move Isaac to set locations and if he goes through a door it records the door position and the position it leads to. I found 16 places doors lead to in Sol Sanctum. I would say with about 90% certainty that this is complete. It is possible that a door I found is unreachable in all cases due to obstacles in the way, but there are multiple occurrences of each door. Certain doors are much much more common than others.

Here is the list of places that doors lead to.

SOL SANCTUM:

Save Retreat Glitch can lead to:

Outside
Room 1 Exit
Room 2 Entrance
Room 3 Entrance
Room 3 Exit
Room 4 Exit
Room 5 Entrance
Room 5 Left Exit
Room 5 Right Exit
Room 5.5 Entrance
Room 6 Entrance
Room 7 Entrance
Room 7 Exit
Room 8 Left Exit
Room 8.5 Entrance
Room 9 Entrance

Room 9 was the best door I found originally. I labeled it previously as Sol Sanctum Door 1: 30, 4282.

Normal Retreat Glitch locations:

Room 3 Entrance to Room 1 Entrance
Room 3 Exit to Room 1 Entrance
Room 7 Entrance to Room 1 Entrance
Room 7 Exit to Room 1 Entrance
Room 9 Entrance to Room 1 Entrance
Room 9 Exit to Room 1 Entrance

Sol Sanctum is pretty much solved under the current knowledge as far as I am concerned. Enter room 3, save retreat glitch, find an off map door to room 9, complete the dungeon as normal.

GOMA CAVE:

Area refers to a new area of a room accessed. Area 3 is the only one you would enter twice in a normal path. The Vale cave with the power bread is in fact on the same map. It is just to the right of the entrance, you can see it if you get off map.

Save Retreat Glitch can lead to:

ENTRANCE POINT "GOMA CAVE ENTRANCE"
Goma Cave Outside
Room 2 Entrance
Area 5 Exit (Up by Mars Djinni)
Area 7 Entrance (Last Room)
Vale Room Outside

ENTRANCE POINT BILIBIN SIDE:
?

Normal Retreat Glitch locations:

ENTRANCE POINT "GOMA CAVE ENTRANCE"
Area 2 Entrance to Room 1 Entrance
Area 2 Exit to Room 1 Entrance
Area 3 Entrance to Room 1 Entrance
Area 3 Exit (to area 4) to Room 1 Entrance
Area 4 Entrance to Room 1 Entrance
Area 3 Exit (to area 5) to Room 1 Entrance
Area 5 Entrance to Room 1 Entrance
Area 5 Exit to Room 1 Entrance
Area 7 Entrance to Outside Vault Side
Area 7 Exit to Room 2 Entrance

ENTRANCE POINT BILIBIN SIDE:
Area 6 Exit to 11 9 off map ********
Area 6 Entrance to 11 9 off map ******
Area 5 Exit to Area 7 Exit
Area 5 Entrance to Area 7 Exit
Area 4 Entrance to Area 7 Exit
Area 3 Exit (to area 5) to Area 7 Exit
Area 3 Exit (to area 4) to Area 7 Exit
Area 3 Entrance to Area 7 Exit
Area 2 Exit to Area 7 Exit
Area 2 Entrance to Area 7 Exit
Area 1 Exit to Outside Bilibin Side
Area 1 Entrance to Area 6 Exit

-------------------

BILIBIN CAVE:

Save Retreat Glitch can lead to:

ENTRANCE POINT "BILIBIN CAVE ENTRANCE"
Bilibin Cave Entrance
Room 2 Top Level South
Room 2 Entrance
Outside Growth Entrance
Before Reveal Door (By Elven Rapier)
Inside Reveal Door

ENTRANCE POINT IMIL SIDE (These doors are in the same off-map positions)
Room 1 Exit
Room 2 Top Level North
Outside Imil Side
Reveal Room North Side
(No Corresponding Door Here?)
Inside Treasure room (Water of Life, Vial)

Normal Retreat Glitch locations:

ENTRANCE POINT "BILIBIN CAVE ENTRANCE"
Room 2 Entrance to Bilibin Cave Entrance
Room 2 Exit to Room 2 Entrance

ENTRANCE POINT IMIL SIDE
Room 1 Exit to Outside Imil Side
Room 1 Entrance to Room 1 Exit

I wrote a bot to move Isaac to set locations and if he goes through a door it records the door position and the position it leads to.  I found 16 places doors lead to in Sol Sanctum.  I would say with about 90% certainty that this is complete.  It is possible that a door I found is unreachable in all cases due to obstacles in the way, but there are multiple occurrences of each door.  Certain doors are much much more common than others.

Here is the list of places that doors lead to.

SOL SANCTUM:

Save Retreat Glitch can lead to:

Outside
Room 1 Exit
Room 2 Entrance
Room 3 Entrance
Room 3 Exit
Room 4 Exit
Room 5 Entrance
Room 5 Left Exit
Room 5 Right Exit
Room 5.5 Entrance
Room 6 Entrance
Room 7 Entrance
Room 7 Exit 
Room 8 Left Exit
Room 8.5 Entrance
Room 9 Entrance

Room 9 was the best door I found originally.  I labeled it previously as Sol Sanctum Door 1: 30, 4282.

Normal Retreat Glitch locations:

Room 3 Entrance to Room 1 Entrance
Room 3 Exit to Room 1 Entrance
Room 7 Entrance to Room 1 Entrance
Room 7 Exit to Room 1 Entrance
Room 9 Entrance to Room 1 Entrance
Room 9 Exit to Room 1 Entrance

Sol Sanctum is pretty much solved under the current knowledge as far as I am concerned.  Enter room 3, save retreat glitch, find an off map door to room 9, complete the dungeon as normal.

GOMA CAVE:

Area refers to a new area of a room accessed. Area 3 is the only one you would enter twice in a normal path.  The Vale cave with the power bread is in fact on the same map.  It is just to the right of the entrance, you can see it if you get off map.

Save Retreat Glitch can lead to:

ENTRANCE POINT "GOMA CAVE ENTRANCE"
Goma Cave Outside
Room 2 Entrance
Area 5 Exit (Up by Mars Djinni)
Area 7 Entrance (Last Room)
Vale Room Outside

ENTRANCE POINT BILIBIN SIDE:
?

Normal Retreat Glitch locations:

ENTRANCE POINT "GOMA CAVE ENTRANCE"
Area 2 Entrance to Room 1 Entrance
Area 2 Exit to Room 1 Entrance
Area 3 Entrance to Room 1 Entrance
Area 3 Exit (to area 4) to Room 1 Entrance
Area 4 Entrance to Room 1 Entrance
Area 3 Exit (to area 5) to Room 1 Entrance
Area 5 Entrance to Room 1 Entrance
Area 5 Exit to Room 1 Entrance
Area 7 Entrance to Outside Vault Side
Area 7 Exit to Room 2 Entrance

ENTRANCE POINT BILIBIN SIDE:
Area 6 Exit to 11 9 off map ********
Area 6 Entrance to 11 9 off map ******
Area 5 Exit to Area 7 Exit
Area 5 Entrance to Area 7 Exit
Area 4 Entrance to Area 7 Exit
Area 3 Exit (to area 5) to Area 7 Exit
Area 3 Exit (to area 4) to Area 7 Exit
Area 3 Entrance to Area 7 Exit
Area 2 Exit to Area 7 Exit
Area 2 Entrance to Area 7 Exit
Area 1 Exit to Outside Bilibin Side
Area 1 Entrance to Area 6 Exit

-------------------

BILIBIN CAVE:

Save Retreat Glitch can lead to:

ENTRANCE POINT "BILIBIN CAVE ENTRANCE"
Bilibin Cave Entrance
Room 2 Top Level South
Room 2 Entrance
Outside Growth Entrance
Before Reveal Door (By Elven Rapier)
Inside Reveal Door

ENTRANCE POINT IMIL SIDE (These doors are in the same off-map positions)
Room 1 Exit
Room 2 Top Level North
Outside Imil Side
Reveal Room North Side
(No Corresponding Door Here?)
Inside Treasure room (Water of Life, Vial)

Normal Retreat Glitch locations:

ENTRANCE POINT "BILIBIN CAVE ENTRANCE"
Room 2 Entrance to Bilibin Cave Entrance
Room 2 Exit to Room 2 Entrance

ENTRANCE POINT IMIL SIDE
Room 1 Exit to Outside Imil Side
Room 1 Entrance to Room 1 Exit